net/ipv6/ip6_tunnel.c

   1 // SPDX-License-Identifier: GPL-2.0-or-later
   2 /*
   3  *      IPv6 tunneling device
   4  *      Linux INET6 implementation
   5  *
   6  *      Authors:
   7  *      Ville Nuorvala          <vnuorval@tcs.hut.fi>
   8  *      Yasuyuki Kozakai        <kozakai@linux-ipv6.org>
   9  *
  10  *      Based on:
  11  *      linux/net/ipv6/sit.c and linux/net/ipv4/ipip.c
  12  *
  13  *      RFC 2473
  14  */
  15
  16 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  17
  18 #include <linux/module.h>
  19 #include <linux/capability.h>
  20 #include <linux/errno.h>
  21 #include <linux/types.h>
  22 #include <linux/sockios.h>
  23 #include <linux/icmp.h>
  24 #include <linux/if.h>
  25 #include <linux/in.h>
  26 #include <linux/ip.h>
  27 #include <linux/net.h>
  28 #include <linux/in6.h>
  29 #include <linux/netdevice.h>
  30 #include <linux/if_arp.h>
  31 #include <linux/icmpv6.h>
  32 #include <linux/init.h>
  33 #include <linux/route.h>
  34 #include <linux/rtnetlink.h>
  35 #include <linux/netfilter_ipv6.h>
  36 #include <linux/slab.h>
  37 #include <linux/hash.h>
  38 #include <linux/etherdevice.h>
  39
  40 #include <linux/uaccess.h>
  41 #include <linux/atomic.h>
  42
  43 #include <net/icmp.h>
  44 #include <net/ip.h>
  45 #include <net/ip_tunnels.h>
  46 #include <net/ipv6.h>
  47 #include <net/ip6_route.h>
  48 #include <net/addrconf.h>
  49 #include <net/ip6_tunnel.h>
  50 #include <net/xfrm.h>
  51 #include <net/dsfield.h>
  52 #include <net/inet_ecn.h>
  53 #include <net/net_namespace.h>
  54 #include <net/netns/generic.h>
  55 #include <net/dst_metadata.h>
  56
  57 MODULE_AUTHOR("Ville Nuorvala");
  58 MODULE_DESCRIPTION("IPv6 tunneling device");
  59 MODULE_LICENSE("GPL");
  60 MODULE_ALIAS_RTNL_LINK("ip6tnl");
  61 MODULE_ALIAS_NETDEV("ip6tnl0");
  62
  63 #define IP6_TUNNEL_HASH_SIZE_SHIFT  5
  64 #define IP6_TUNNEL_HASH_SIZE (1 << IP6_TUNNEL_HASH_SIZE_SHIFT)
  65
  66 static bool log_ecn_error = true;
  67 module_param(log_ecn_error, bool, 0644);
  68 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
  69
  70 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2)
  71 {
  72         u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2);
  73
  74         return hash_32(hash, IP6_TUNNEL_HASH_SIZE_SHIFT);
  75 }
  76
  77 static int ip6_tnl_dev_init(struct net_device *dev);
  78 static void ip6_tnl_dev_setup(struct net_device *dev);
  79 static struct rtnl_link_ops ip6_link_ops __read_mostly;
  80
  81 static unsigned int ip6_tnl_net_id __read_mostly;
  82 struct ip6_tnl_net {
  83         /* the IPv6 tunnel fallback device */
  84         struct net_device *fb_tnl_dev;
  85         /* lists for storing tunnels in use */
  86         struct ip6_tnl __rcu *tnls_r_l[IP6_TUNNEL_HASH_SIZE];
  87         struct ip6_tnl __rcu *tnls_wc[1];
  88         struct ip6_tnl __rcu **tnls[2];
  89         struct ip6_tnl __rcu *collect_md_tun;
  90 };
  91
  92 static struct net_device_stats *ip6_get_stats(struct net_device *dev)
  93 {
  94         struct pcpu_sw_netstats tmp, sum = { 0 };
  95         int i;
  96
  97         for_each_possible_cpu(i) {
  98                 unsigned int start;
  99                 const struct pcpu_sw_netstats *tstats =
 100                                                    per_cpu_ptr(dev->tstats, i);
 101
 102                 do {
 103                         start = u64_stats_fetch_begin_irq(&tstats->syncp);
 104                         tmp.rx_packets = tstats->rx_packets;
 105                         tmp.rx_bytes = tstats->rx_bytes;
 106                         tmp.tx_packets = tstats->tx_packets;
 107                         tmp.tx_bytes =  tstats->tx_bytes;
 108                 } while (u64_stats_fetch_retry_irq(&tstats->syncp, start));
 109
 110                 sum.rx_packets += tmp.rx_packets;
 111                 sum.rx_bytes   += tmp.rx_bytes;
 112                 sum.tx_packets += tmp.tx_packets;
 113                 sum.tx_bytes   += tmp.tx_bytes;
 114         }
 115         dev->stats.rx_packets = sum.rx_packets;
 116         dev->stats.rx_bytes   = sum.rx_bytes;
 117         dev->stats.tx_packets = sum.tx_packets;
 118         dev->stats.tx_bytes   = sum.tx_bytes;
 119         return &dev->stats;
 120 }
 121
 122 /**
 123  * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
 124  *   @remote: the address of the tunnel exit-point
 125  *   @local: the address of the tunnel entry-point
 126  *
 127  * Return:
 128  *   tunnel matching given end-points if found,
 129  *   else fallback tunnel if its device is up,
 130  *   else %NULL
 131  **/
 132
 133 #define for_each_ip6_tunnel_rcu(start) \
 134         for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
 135
 136 static struct ip6_tnl *
 137 ip6_tnl_lookup(struct net *net, const struct in6_addr *remote, const struct in6_addr *local)
 138 {
 139         unsigned int hash = HASH(remote, local);
 140         struct ip6_tnl *t;
 141         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 142         struct in6_addr any;
 143
 144         for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
 145                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
 146                     ipv6_addr_equal(remote, &t->parms.raddr) &&
 147                     (t->dev->flags & IFF_UP))
 148                         return t;
 149         }
 150
 151         memset(&any, 0, sizeof(any));
 152         hash = HASH(&any, local);
 153         for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
 154                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
 155                     ipv6_addr_any(&t->parms.raddr) &&
 156                     (t->dev->flags & IFF_UP))
 157                         return t;
 158         }
 159
 160         hash = HASH(remote, &any);
 161         for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
 162                 if (ipv6_addr_equal(remote, &t->parms.raddr) &&
 163                     ipv6_addr_any(&t->parms.laddr) &&
 164                     (t->dev->flags & IFF_UP))
 165                         return t;
 166         }
 167
 168         t = rcu_dereference(ip6n->collect_md_tun);
 169         if (t && t->dev->flags & IFF_UP)
 170                 return t;
 171
 172         t = rcu_dereference(ip6n->tnls_wc[0]);
 173         if (t && (t->dev->flags & IFF_UP))
 174                 return t;
 175
 176         return NULL;
 177 }
 178
 179 /**
 180  * ip6_tnl_bucket - get head of list matching given tunnel parameters
 181  *   @p: parameters containing tunnel end-points
 182  *
 183  * Description:
 184  *   ip6_tnl_bucket() returns the head of the list matching the
 185  *   &struct in6_addr entries laddr and raddr in @p.
 186  *
 187  * Return: head of IPv6 tunnel list
 188  **/
 189
 190 static struct ip6_tnl __rcu **
 191 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, const struct __ip6_tnl_parm *p)
 192 {
 193         const struct in6_addr *remote = &p->raddr;
 194         const struct in6_addr *local = &p->laddr;
 195         unsigned int h = 0;
 196         int prio = 0;
 197
 198         if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
 199                 prio = 1;
 200                 h = HASH(remote, local);
 201         }
 202         return &ip6n->tnls[prio][h];
 203 }
 204
 205 /**
 206  * ip6_tnl_link - add tunnel to hash table
 207  *   @t: tunnel to be added
 208  **/
 209
 210 static void
 211 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
 212 {
 213         struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms);
 214
 215         if (t->parms.collect_md)
 216                 rcu_assign_pointer(ip6n->collect_md_tun, t);
 217         rcu_assign_pointer(t->next , rtnl_dereference(*tp));
 218         rcu_assign_pointer(*tp, t);
 219 }
 220
 221 /**
 222  * ip6_tnl_unlink - remove tunnel from hash table
 223  *   @t: tunnel to be removed
 224  **/
 225
 226 static void
 227 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
 228 {
 229         struct ip6_tnl __rcu **tp;
 230         struct ip6_tnl *iter;
 231
 232         if (t->parms.collect_md)
 233                 rcu_assign_pointer(ip6n->collect_md_tun, NULL);
 234
 235         for (tp = ip6_tnl_bucket(ip6n, &t->parms);
 236              (iter = rtnl_dereference(*tp)) != NULL;
 237              tp = &iter->next) {
 238                 if (t == iter) {
 239                         rcu_assign_pointer(*tp, t->next);
 240                         break;
 241                 }
 242         }
 243 }
 244
 245 static void ip6_dev_free(struct net_device *dev)
 246 {
 247         struct ip6_tnl *t = netdev_priv(dev);
 248
 249         gro_cells_destroy(&t->gro_cells);
 250         dst_cache_destroy(&t->dst_cache);
 251         free_percpu(dev->tstats);
 252 }
 253
 254 static int ip6_tnl_create2(struct net_device *dev)
 255 {
 256         struct ip6_tnl *t = netdev_priv(dev);
 257         struct net *net = dev_net(dev);
 258         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 259         int err;
 260
 261         t = netdev_priv(dev);
 262
 263         dev->rtnl_link_ops = &ip6_link_ops;
 264         err = register_netdevice(dev);
 265         if (err < 0)
 266                 goto out;
 267
 268         strcpy(t->parms.name, dev->name);
 269
 270         dev_hold(dev);
 271         ip6_tnl_link(ip6n, t);
 272         return 0;
 273
 274 out:
 275         return err;
 276 }
 277
 278 /**
 279  * ip6_tnl_create - create a new tunnel
 280  *   @p: tunnel parameters
 281  *   @pt: pointer to new tunnel
 282  *
 283  * Description:
 284  *   Create tunnel matching given parameters.
 285  *
 286  * Return:
 287  *   created tunnel or error pointer
 288  **/
 289
 290 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
 291 {
 292         struct net_device *dev;
 293         struct ip6_tnl *t;
 294         char name[IFNAMSIZ];
 295         int err = -E2BIG;
 296
 297         if (p->name[0]) {
 298                 if (!dev_valid_name(p->name))
 299                         goto failed;
 300                 strlcpy(name, p->name, IFNAMSIZ);
 301         } else {
 302                 sprintf(name, "ip6tnl%%d");
 303         }
 304         err = -ENOMEM;
 305         dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
 306                            ip6_tnl_dev_setup);
 307         if (!dev)
 308                 goto failed;
 309
 310         dev_net_set(dev, net);
 311
 312         t = netdev_priv(dev);
 313         t->parms = *p;
 314         t->net = dev_net(dev);
 315         err = ip6_tnl_create2(dev);
 316         if (err < 0)
 317                 goto failed_free;
 318
 319         return t;
 320
 321 failed_free:
 322         free_netdev(dev);
 323 failed:
 324         return ERR_PTR(err);
 325 }
 326
 327 /**
 328  * ip6_tnl_locate - find or create tunnel matching given parameters
 329  *   @p: tunnel parameters
 330  *   @create: != 0 if allowed to create new tunnel if no match found
 331  *
 332  * Description:
 333  *   ip6_tnl_locate() first tries to locate an existing tunnel
 334  *   based on @parms. If this is unsuccessful, but @create is set a new
 335  *   tunnel device is created and registered for use.
 336  *
 337  * Return:
 338  *   matching tunnel or error pointer
 339  **/
 340
 341 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
 342                 struct __ip6_tnl_parm *p, int create)
 343 {
 344         const struct in6_addr *remote = &p->raddr;
 345         const struct in6_addr *local = &p->laddr;
 346         struct ip6_tnl __rcu **tp;
 347         struct ip6_tnl *t;
 348         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 349
 350         for (tp = ip6_tnl_bucket(ip6n, p);
 351              (t = rtnl_dereference(*tp)) != NULL;
 352              tp = &t->next) {
 353                 if (ipv6_addr_equal(local, &t->parms.laddr) &&
 354                     ipv6_addr_equal(remote, &t->parms.raddr)) {
 355                         if (create)
 356                                 return ERR_PTR(-EEXIST);
 357
 358                         return t;
 359                 }
 360         }
 361         if (!create)
 362                 return ERR_PTR(-ENODEV);
 363         return ip6_tnl_create(net, p);
 364 }
 365
 366 /**
 367  * ip6_tnl_dev_uninit - tunnel device uninitializer
 368  *   @dev: the device to be destroyed
 369  *
 370  * Description:
 371  *   ip6_tnl_dev_uninit() removes tunnel from its list
 372  **/
 373
 374 static void
 375 ip6_tnl_dev_uninit(struct net_device *dev)
 376 {
 377         struct ip6_tnl *t = netdev_priv(dev);
 378         struct net *net = t->net;
 379         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 380
 381         if (dev == ip6n->fb_tnl_dev)
 382                 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
 383         else
 384                 ip6_tnl_unlink(ip6n, t);
 385         dst_cache_reset(&t->dst_cache);
 386         dev_put(dev);
 387 }
 388
 389 /**
 390  * parse_tvl_tnl_enc_lim - handle encapsulation limit option
 391  *   @skb: received socket buffer
 392  *
 393  * Return:
 394  *   0 if none was found,
 395  *   else index to encapsulation limit
 396  **/
 397
 398 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw)
 399 {
 400         const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)raw;
 401         unsigned int nhoff = raw - skb->data;
 402         unsigned int off = nhoff + sizeof(*ipv6h);
 403         u8 next, nexthdr = ipv6h->nexthdr;
 404
 405         while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
 406                 struct ipv6_opt_hdr *hdr;
 407                 u16 optlen;
 408
 409                 if (!pskb_may_pull(skb, off + sizeof(*hdr)))
 410                         break;
 411
 412                 hdr = (struct ipv6_opt_hdr *)(skb->data + off);
 413                 if (nexthdr == NEXTHDR_FRAGMENT) {
 414                         struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
 415                         if (frag_hdr->frag_off)
 416                                 break;
 417                         optlen = 8;
 418                 } else if (nexthdr == NEXTHDR_AUTH) {
 419                         optlen = (hdr->hdrlen + 2) << 2;
 420                 } else {
 421                         optlen = ipv6_optlen(hdr);
 422                 }
 423                 /* cache hdr->nexthdr, since pskb_may_pull() might
 424                  * invalidate hdr
 425                  */
 426                 next = hdr->nexthdr;
 427                 if (nexthdr == NEXTHDR_DEST) {
 428                         u16 i = 2;
 429
 430                         /* Remember : hdr is no longer valid at this point. */
 431                         if (!pskb_may_pull(skb, off + optlen))
 432                                 break;
 433
 434                         while (1) {
 435                                 struct ipv6_tlv_tnl_enc_lim *tel;
 436
 437                                 /* No more room for encapsulation limit */
 438                                 if (i + sizeof(*tel) > optlen)
 439                                         break;
 440
 441                                 tel = (struct ipv6_tlv_tnl_enc_lim *)(skb->data + off + i);
 442                                 /* return index of option if found and valid */
 443                                 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
 444                                     tel->length == 1)
 445                                         return i + off - nhoff;
 446                                 /* else jump to next option */
 447                                 if (tel->type)
 448                                         i += tel->length + 2;
 449                                 else
 450                                         i++;
 451                         }
 452                 }
 453                 nexthdr = next;
 454                 off += optlen;
 455         }
 456         return 0;
 457 }
 458 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim);
 459
 460 /**
 461  * ip6_tnl_err - tunnel error handler
 462  *
 463  * Description:
 464  *   ip6_tnl_err() should handle errors in the tunnel according
 465  *   to the specifications in RFC 2473.
 466  **/
 467
 468 static int
 469 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
 470             u8 *type, u8 *code, int *msg, __u32 *info, int offset)
 471 {
 472         const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)skb->data;
 473         struct net *net = dev_net(skb->dev);
 474         u8 rel_type = ICMPV6_DEST_UNREACH;
 475         u8 rel_code = ICMPV6_ADDR_UNREACH;
 476         __u32 rel_info = 0;
 477         struct ip6_tnl *t;
 478         int err = -ENOENT;
 479         int rel_msg = 0;
 480         u8 tproto;
 481         __u16 len;
 482
 483         /* If the packet doesn't contain the original IPv6 header we are
 484            in trouble since we might need the source address for further
 485            processing of the error. */
 486
 487         rcu_read_lock();
 488         t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr, &ipv6h->saddr);
 489         if (!t)
 490                 goto out;
 491
 492         tproto = READ_ONCE(t->parms.proto);
 493         if (tproto != ipproto && tproto != 0)
 494                 goto out;
 495
 496         err = 0;
 497
 498         switch (*type) {
 499                 struct ipv6_tlv_tnl_enc_lim *tel;
 500                 __u32 mtu, teli;
 501         case ICMPV6_DEST_UNREACH:
 502                 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
 503                                     t->parms.name);
 504                 rel_msg = 1;
 505                 break;
 506         case ICMPV6_TIME_EXCEED:
 507                 if ((*code) == ICMPV6_EXC_HOPLIMIT) {
 508                         net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
 509                                             t->parms.name);
 510                         rel_msg = 1;
 511                 }
 512                 break;
 513         case ICMPV6_PARAMPROB:
 514                 teli = 0;
 515                 if ((*code) == ICMPV6_HDR_FIELD)
 516                         teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
 517
 518                 if (teli && teli == *info - 2) {
 519                         tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
 520                         if (tel->encap_limit == 0) {
 521                                 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
 522                                                     t->parms.name);
 523                                 rel_msg = 1;
 524                         }
 525                 } else {
 526                         net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
 527                                             t->parms.name);
 528                 }
 529                 break;
 530         case ICMPV6_PKT_TOOBIG:
 531                 ip6_update_pmtu(skb, net, htonl(*info), 0, 0,
 532                                 sock_net_uid(net, NULL));
 533                 mtu = *info - offset;
 534                 if (mtu < IPV6_MIN_MTU)
 535                         mtu = IPV6_MIN_MTU;
 536                 len = sizeof(*ipv6h) + ntohs(ipv6h->payload_len);
 537                 if (len > mtu) {
 538                         rel_type = ICMPV6_PKT_TOOBIG;
 539                         rel_code = 0;
 540                         rel_info = mtu;
 541                         rel_msg = 1;
 542                 }
 543                 break;
 544         case NDISC_REDIRECT:
 545                 ip6_redirect(skb, net, skb->dev->ifindex, 0,
 546                              sock_net_uid(net, NULL));
 547                 break;
 548         }
 549
 550         *type = rel_type;
 551         *code = rel_code;
 552         *info = rel_info;
 553         *msg = rel_msg;
 554
 555 out:
 556         rcu_read_unlock();
 557         return err;
 558 }
 559
 560 static int
 561 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 562            u8 type, u8 code, int offset, __be32 info)
 563 {
 564         __u32 rel_info = ntohl(info);
 565         const struct iphdr *eiph;
 566         struct sk_buff *skb2;
 567         int err, rel_msg = 0;
 568         u8 rel_type = type;
 569         u8 rel_code = code;
 570         struct rtable *rt;
 571         struct flowi4 fl4;
 572
 573         err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
 574                           &rel_msg, &rel_info, offset);
 575         if (err < 0)
 576                 return err;
 577
 578         if (rel_msg == 0)
 579                 return 0;
 580
 581         switch (rel_type) {
 582         case ICMPV6_DEST_UNREACH:
 583                 if (rel_code != ICMPV6_ADDR_UNREACH)
 584                         return 0;
 585                 rel_type = ICMP_DEST_UNREACH;
 586                 rel_code = ICMP_HOST_UNREACH;
 587                 break;
 588         case ICMPV6_PKT_TOOBIG:
 589                 if (rel_code != 0)
 590                         return 0;
 591                 rel_type = ICMP_DEST_UNREACH;
 592                 rel_code = ICMP_FRAG_NEEDED;
 593                 break;
 594         default:
 595                 return 0;
 596         }
 597
 598         if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
 599                 return 0;
 600
 601         skb2 = skb_clone(skb, GFP_ATOMIC);
 602         if (!skb2)
 603                 return 0;
 604
 605         skb_dst_drop(skb2);
 606
 607         skb_pull(skb2, offset);
 608         skb_reset_network_header(skb2);
 609         eiph = ip_hdr(skb2);
 610
 611         /* Try to guess incoming interface */
 612         rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL, eiph->saddr,
 613                                    0, 0, 0, IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
 614         if (IS_ERR(rt))
 615                 goto out;
 616
 617         skb2->dev = rt->dst.dev;
 618         ip_rt_put(rt);
 619
 620         /* route "incoming" packet */
 621         if (rt->rt_flags & RTCF_LOCAL) {
 622                 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
 623                                            eiph->daddr, eiph->saddr, 0, 0,
 624                                            IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
 625                 if (IS_ERR(rt) || rt->dst.dev->type != ARPHRD_TUNNEL6) {
 626                         if (!IS_ERR(rt))
 627                                 ip_rt_put(rt);
 628                         goto out;
 629                 }
 630                 skb_dst_set(skb2, &rt->dst);
 631         } else {
 632                 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
 633                                    skb2->dev) ||
 634                     skb_dst(skb2)->dev->type != ARPHRD_TUNNEL6)
 635                         goto out;
 636         }
 637
 638         /* change mtu on this route */
 639         if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
 640                 if (rel_info > dst_mtu(skb_dst(skb2)))
 641                         goto out;
 642
 643                 skb_dst_update_pmtu(skb2, rel_info);
 644         }
 645
 646         icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
 647
 648 out:
 649         kfree_skb(skb2);
 650         return 0;
 651 }
 652
 653 static int
 654 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 655            u8 type, u8 code, int offset, __be32 info)
 656 {
 657         __u32 rel_info = ntohl(info);
 658         int err, rel_msg = 0;
 659         u8 rel_type = type;
 660         u8 rel_code = code;
 661
 662         err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
 663                           &rel_msg, &rel_info, offset);
 664         if (err < 0)
 665                 return err;
 666
 667         if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
 668                 struct rt6_info *rt;
 669                 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
 670
 671                 if (!skb2)
 672                         return 0;
 673
 674                 skb_dst_drop(skb2);
 675                 skb_pull(skb2, offset);
 676                 skb_reset_network_header(skb2);
 677
 678                 /* Try to guess incoming interface */
 679                 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
 680                                 NULL, 0, skb2, 0);
 681
 682                 if (rt && rt->dst.dev)
 683                         skb2->dev = rt->dst.dev;
 684
 685                 icmpv6_send(skb2, rel_type, rel_code, rel_info);
 686
 687                 ip6_rt_put(rt);
 688
 689                 kfree_skb(skb2);
 690         }
 691
 692         return 0;
 693 }
 694
 695 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
 696                                        const struct ipv6hdr *ipv6h,
 697                                        struct sk_buff *skb)
 698 {
 699         __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
 700
 701         if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
 702                 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
 703
 704         return IP6_ECN_decapsulate(ipv6h, skb);
 705 }
 706
 707 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
 708                                        const struct ipv6hdr *ipv6h,
 709                                        struct sk_buff *skb)
 710 {
 711         if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
 712                 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
 713
 714         return IP6_ECN_decapsulate(ipv6h, skb);
 715 }
 716
 717 __u32 ip6_tnl_get_cap(struct ip6_tnl *t,
 718                              const struct in6_addr *laddr,
 719                              const struct in6_addr *raddr)
 720 {
 721         struct __ip6_tnl_parm *p = &t->parms;
 722         int ltype = ipv6_addr_type(laddr);
 723         int rtype = ipv6_addr_type(raddr);
 724         __u32 flags = 0;
 725
 726         if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) {
 727                 flags = IP6_TNL_F_CAP_PER_PACKET;
 728         } else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
 729                    rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
 730                    !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
 731                    (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
 732                 if (ltype&IPV6_ADDR_UNICAST)
 733                         flags |= IP6_TNL_F_CAP_XMIT;
 734                 if (rtype&IPV6_ADDR_UNICAST)
 735                         flags |= IP6_TNL_F_CAP_RCV;
 736         }
 737         return flags;
 738 }
 739 EXPORT_SYMBOL(ip6_tnl_get_cap);
 740
 741 /* called with rcu_read_lock() */
 742 int ip6_tnl_rcv_ctl(struct ip6_tnl *t,
 743                                   const struct in6_addr *laddr,
 744                                   const struct in6_addr *raddr)
 745 {
 746         struct __ip6_tnl_parm *p = &t->parms;
 747         int ret = 0;
 748         struct net *net = t->net;
 749
 750         if ((p->flags & IP6_TNL_F_CAP_RCV) ||
 751             ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
 752              (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) {
 753                 struct net_device *ldev = NULL;
 754
 755                 if (p->link)
 756                         ldev = dev_get_by_index_rcu(net, p->link);
 757
 758                 if ((ipv6_addr_is_multicast(laddr) ||
 759                      likely(ipv6_chk_addr_and_flags(net, laddr, ldev, false,
 760                                                     0, IFA_F_TENTATIVE))) &&
 761                     ((p->flags & IP6_TNL_F_ALLOW_LOCAL_REMOTE) ||
 762                      likely(!ipv6_chk_addr_and_flags(net, raddr, ldev, true,
 763                                                      0, IFA_F_TENTATIVE))))
 764                         ret = 1;
 765         }
 766         return ret;
 767 }
 768 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
 769
 770 static int __ip6_tnl_rcv(struct ip6_tnl *tunnel, struct sk_buff *skb,
 771                          const struct tnl_ptk_info *tpi,
 772                          struct metadata_dst *tun_dst,
 773                          int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
 774                                                 const struct ipv6hdr *ipv6h,
 775                                                 struct sk_buff *skb),
 776                          bool log_ecn_err)
 777 {
 778         struct pcpu_sw_netstats *tstats;
 779         const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 780         int err;
 781
 782         if ((!(tpi->flags & TUNNEL_CSUM) &&
 783              (tunnel->parms.i_flags & TUNNEL_CSUM)) ||
 784             ((tpi->flags & TUNNEL_CSUM) &&
 785              !(tunnel->parms.i_flags & TUNNEL_CSUM))) {
 786                 tunnel->dev->stats.rx_crc_errors++;
 787                 tunnel->dev->stats.rx_errors++;
 788                 goto drop;
 789         }
 790
 791         if (tunnel->parms.i_flags & TUNNEL_SEQ) {
 792                 if (!(tpi->flags & TUNNEL_SEQ) ||
 793                     (tunnel->i_seqno &&
 794                      (s32)(ntohl(tpi->seq) - tunnel->i_seqno) < 0)) {
 795                         tunnel->dev->stats.rx_fifo_errors++;
 796                         tunnel->dev->stats.rx_errors++;
 797                         goto drop;
 798                 }
 799                 tunnel->i_seqno = ntohl(tpi->seq) + 1;
 800         }
 801
 802         skb->protocol = tpi->proto;
 803
 804         /* Warning: All skb pointers will be invalidated! */
 805         if (tunnel->dev->type == ARPHRD_ETHER) {
 806                 if (!pskb_may_pull(skb, ETH_HLEN)) {
 807                         tunnel->dev->stats.rx_length_errors++;
 808                         tunnel->dev->stats.rx_errors++;
 809                         goto drop;
 810                 }
 811
 812                 ipv6h = ipv6_hdr(skb);
 813                 skb->protocol = eth_type_trans(skb, tunnel->dev);
 814                 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
 815         } else {
 816                 skb->dev = tunnel->dev;
 817         }
 818
 819         skb_reset_network_header(skb);
 820         memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
 821
 822         __skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
 823
 824         err = dscp_ecn_decapsulate(tunnel, ipv6h, skb);
 825         if (unlikely(err)) {
 826                 if (log_ecn_err)
 827                         net_info_ratelimited("non-ECT from %pI6 with DS=%#x\n",
 828                                              &ipv6h->saddr,
 829                                              ipv6_get_dsfield(ipv6h));
 830                 if (err > 1) {
 831                         ++tunnel->dev->stats.rx_frame_errors;
 832                         ++tunnel->dev->stats.rx_errors;
 833                         goto drop;
 834                 }
 835         }
 836
 837         tstats = this_cpu_ptr(tunnel->dev->tstats);
 838         u64_stats_update_begin(&tstats->syncp);
 839         tstats->rx_packets++;
 840         tstats->rx_bytes += skb->len;
 841         u64_stats_update_end(&tstats->syncp);
 842
 843         skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(tunnel->dev)));
 844
 845         if (tun_dst)
 846                 skb_dst_set(skb, (struct dst_entry *)tun_dst);
 847
 848         gro_cells_receive(&tunnel->gro_cells, skb);
 849         return 0;
 850
 851 drop:
 852         if (tun_dst)
 853                 dst_release((struct dst_entry *)tun_dst);
 854         kfree_skb(skb);
 855         return 0;
 856 }
 857
 858 int ip6_tnl_rcv(struct ip6_tnl *t, struct sk_buff *skb,
 859                 const struct tnl_ptk_info *tpi,
 860                 struct metadata_dst *tun_dst,
 861                 bool log_ecn_err)
 862 {
 863         return __ip6_tnl_rcv(t, skb, tpi, tun_dst, ip6ip6_dscp_ecn_decapsulate,
 864                              log_ecn_err);
 865 }
 866 EXPORT_SYMBOL(ip6_tnl_rcv);
 867
 868 static const struct tnl_ptk_info tpi_v6 = {
 869         /* no tunnel info required for ipxip6. */
 870         .proto = htons(ETH_P_IPV6),
 871 };
 872
 873 static const struct tnl_ptk_info tpi_v4 = {
 874         /* no tunnel info required for ipxip6. */
 875         .proto = htons(ETH_P_IP),
 876 };
 877
 878 static int ipxip6_rcv(struct sk_buff *skb, u8 ipproto,
 879                       const struct tnl_ptk_info *tpi,
 880                       int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
 881                                                   const struct ipv6hdr *ipv6h,
 882                                                   struct sk_buff *skb))
 883 {
 884         struct ip6_tnl *t;
 885         const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 886         struct metadata_dst *tun_dst = NULL;
 887         int ret = -1;
 888
 889         rcu_read_lock();
 890         t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr);
 891
 892         if (t) {
 893                 u8 tproto = READ_ONCE(t->parms.proto);
 894
 895                 if (tproto != ipproto && tproto != 0)
 896                         goto drop;
 897                 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
 898                         goto drop;
 899                 ipv6h = ipv6_hdr(skb);
 900                 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr))
 901                         goto drop;
 902                 if (iptunnel_pull_header(skb, 0, tpi->proto, false))
 903                         goto drop;
 904                 if (t->parms.collect_md) {
 905                         tun_dst = ipv6_tun_rx_dst(skb, 0, 0, 0);
 906                         if (!tun_dst)
 907                                 goto drop;
 908                 }
 909                 ret = __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate,
 910                                     log_ecn_error);
 911         }
 912
 913         rcu_read_unlock();
 914
 915         return ret;
 916
 917 drop:
 918         rcu_read_unlock();
 919         kfree_skb(skb);
 920         return 0;
 921 }
 922
 923 static int ip4ip6_rcv(struct sk_buff *skb)
 924 {
 925         return ipxip6_rcv(skb, IPPROTO_IPIP, &tpi_v4,
 926                           ip4ip6_dscp_ecn_decapsulate);
 927 }
 928
 929 static int ip6ip6_rcv(struct sk_buff *skb)
 930 {
 931         return ipxip6_rcv(skb, IPPROTO_IPV6, &tpi_v6,
 932                           ip6ip6_dscp_ecn_decapsulate);
 933 }
 934
 935 struct ipv6_tel_txoption {
 936         struct ipv6_txoptions ops;
 937         __u8 dst_opt[8];
 938 };
 939
 940 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
 941 {
 942         memset(opt, 0, sizeof(struct ipv6_tel_txoption));
 943
 944         opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
 945         opt->dst_opt[3] = 1;
 946         opt->dst_opt[4] = encap_limit;
 947         opt->dst_opt[5] = IPV6_TLV_PADN;
 948         opt->dst_opt[6] = 1;
 949
 950         opt->ops.dst1opt = (struct ipv6_opt_hdr *) opt->dst_opt;
 951         opt->ops.opt_nflen = 8;
 952 }
 953
 954 /**
 955  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
 956  *   @t: the outgoing tunnel device
 957  *   @hdr: IPv6 header from the incoming packet
 958  *
 959  * Description:
 960  *   Avoid trivial tunneling loop by checking that tunnel exit-point
 961  *   doesn't match source of incoming packet.
 962  *
 963  * Return:
 964  *   1 if conflict,
 965  *   0 else
 966  **/
 967
 968 static inline bool
 969 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
 970 {
 971         return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
 972 }
 973
 974 int ip6_tnl_xmit_ctl(struct ip6_tnl *t,
 975                      const struct in6_addr *laddr,
 976                      const struct in6_addr *raddr)
 977 {
 978         struct __ip6_tnl_parm *p = &t->parms;
 979         int ret = 0;
 980         struct net *net = t->net;
 981
 982         if (t->parms.collect_md)
 983                 return 1;
 984
 985         if ((p->flags & IP6_TNL_F_CAP_XMIT) ||
 986             ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
 987              (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_XMIT))) {
 988                 struct net_device *ldev = NULL;
 989
 990                 rcu_read_lock();
 991                 if (p->link)
 992                         ldev = dev_get_by_index_rcu(net, p->link);
 993
 994                 if (unlikely(!ipv6_chk_addr_and_flags(net, laddr, ldev, false,
 995                                                       0, IFA_F_TENTATIVE)))
 996                         pr_warn("%s xmit: Local address not yet configured!\n",
 997                                 p->name);
 998                 else if (!(p->flags & IP6_TNL_F_ALLOW_LOCAL_REMOTE) &&
 999                          !ipv6_addr_is_multicast(raddr) &&
1000                          unlikely(ipv6_chk_addr_and_flags(net, raddr, ldev,
1001                                                           true, 0, IFA_F_TENTATIVE)))
1002                         pr_warn("%s xmit: Routing loop! Remote address found on this node!\n",
1003                                 p->name);
1004                 else
1005                         ret = 1;
1006                 rcu_read_unlock();
1007         }
1008         return ret;
1009 }
1010 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
1011
1012 /**
1013  * ip6_tnl_xmit - encapsulate packet and send
1014  *   @skb: the outgoing socket buffer
1015  *   @dev: the outgoing tunnel device
1016  *   @dsfield: dscp code for outer header
1017  *   @fl6: flow of tunneled packet
1018  *   @encap_limit: encapsulation limit
1019  *   @pmtu: Path MTU is stored if packet is too big
1020  *   @proto: next header value
1021  *
1022  * Description:
1023  *   Build new header and do some sanity checks on the packet before sending
1024  *   it.
1025  *
1026  * Return:
1027  *   0 on success
1028  *   -1 fail
1029  *   %-EMSGSIZE message too big. return mtu in this case.
1030  **/
1031
1032 int ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev, __u8 dsfield,
1033                  struct flowi6 *fl6, int encap_limit, __u32 *pmtu,
1034                  __u8 proto)
1035 {
1036         struct ip6_tnl *t = netdev_priv(dev);
1037         struct net *net = t->net;
1038         struct net_device_stats *stats = &t->dev->stats;
1039         struct ipv6hdr *ipv6h;
1040         struct ipv6_tel_txoption opt;
1041         struct dst_entry *dst = NULL, *ndst = NULL;
1042         struct net_device *tdev;
1043         int mtu;
1044         unsigned int eth_hlen = t->dev->type == ARPHRD_ETHER ? ETH_HLEN : 0;
1045         unsigned int psh_hlen = sizeof(struct ipv6hdr) + t->encap_hlen;
1046         unsigned int max_headroom = psh_hlen;
1047         bool use_cache = false;
1048         u8 hop_limit;
1049         int err = -1;
1050
1051         if (t->parms.collect_md) {
1052                 hop_limit = skb_tunnel_info(skb)->key.ttl;
1053                 goto route_lookup;
1054         } else {
1055                 hop_limit = t->parms.hop_limit;
1056         }
1057
1058         /* NBMA tunnel */
1059         if (ipv6_addr_any(&t->parms.raddr)) {
1060                 if (skb->protocol == htons(ETH_P_IPV6)) {
1061                         struct in6_addr *addr6;
1062                         struct neighbour *neigh;
1063                         int addr_type;
1064
1065                         if (!skb_dst(skb))
1066                                 goto tx_err_link_failure;
1067
1068                         neigh = dst_neigh_lookup(skb_dst(skb),
1069                                                  &ipv6_hdr(skb)->daddr);
1070                         if (!neigh)
1071                                 goto tx_err_link_failure;
1072
1073                         addr6 = (struct in6_addr *)&neigh->primary_key;
1074                         addr_type = ipv6_addr_type(addr6);
1075
1076                         if (addr_type == IPV6_ADDR_ANY)
1077                                 addr6 = &ipv6_hdr(skb)->daddr;
1078
1079                         memcpy(&fl6->daddr, addr6, sizeof(fl6->daddr));
1080                         neigh_release(neigh);
1081                 }
1082         } else if (t->parms.proto != 0 && !(t->parms.flags &
1083                                             (IP6_TNL_F_USE_ORIG_TCLASS |
1084                                              IP6_TNL_F_USE_ORIG_FWMARK))) {
1085                 /* enable the cache only if neither the outer protocol nor the
1086                  * routing decision depends on the current inner header value
1087                  */
1088                 use_cache = true;
1089         }
1090
1091         if (use_cache)
1092                 dst = dst_cache_get(&t->dst_cache);
1093
1094         if (!ip6_tnl_xmit_ctl(t, &fl6->saddr, &fl6->daddr))
1095                 goto tx_err_link_failure;
1096
1097         if (!dst) {
1098 route_lookup:
1099                 /* add dsfield to flowlabel for route lookup */
1100                 fl6->flowlabel = ip6_make_flowinfo(dsfield, fl6->flowlabel);
1101
1102                 dst = ip6_route_output(net, NULL, fl6);
1103
1104                 if (dst->error)
1105                         goto tx_err_link_failure;
1106                 dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0);
1107                 if (IS_ERR(dst)) {
1108                         err = PTR_ERR(dst);
1109                         dst = NULL;
1110                         goto tx_err_link_failure;
1111                 }
1112                 if (t->parms.collect_md && ipv6_addr_any(&fl6->saddr) &&
1113                     ipv6_dev_get_saddr(net, ip6_dst_idev(dst)->dev,
1114                                        &fl6->daddr, 0, &fl6->saddr))
1115                         goto tx_err_link_failure;
1116                 ndst = dst;
1117         }
1118
1119         tdev = dst->dev;
1120
1121         if (tdev == dev) {
1122                 stats->collisions++;
1123                 net_warn_ratelimited("%s: Local routing loop detected!\n",
1124                                      t->parms.name);
1125                 goto tx_err_dst_release;
1126         }
1127         mtu = dst_mtu(dst) - eth_hlen - psh_hlen - t->tun_hlen;
1128         if (encap_limit >= 0) {
1129                 max_headroom += 8;
1130                 mtu -= 8;
1131         }
1132         mtu = max(mtu, skb->protocol == htons(ETH_P_IPV6) ?
1133                        IPV6_MIN_MTU : IPV4_MIN_MTU);
1134
1135         skb_dst_update_pmtu(skb, mtu);
1136         if (skb->len - t->tun_hlen - eth_hlen > mtu && !skb_is_gso(skb)) {
1137                 *pmtu = mtu;
1138                 err = -EMSGSIZE;
1139                 goto tx_err_dst_release;
1140         }
1141
1142         if (t->err_count > 0) {
1143                 if (time_before(jiffies,
1144                                 t->err_time + IP6TUNNEL_ERR_TIMEO)) {
1145                         t->err_count--;
1146
1147                         dst_link_failure(skb);
1148                 } else {
1149                         t->err_count = 0;
1150                 }
1151         }
1152
1153         skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
1154
1155         /*
1156          * Okay, now see if we can stuff it in the buffer as-is.
1157          */
1158         max_headroom += LL_RESERVED_SPACE(tdev);
1159
1160         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1161             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1162                 struct sk_buff *new_skb;
1163
1164                 new_skb = skb_realloc_headroom(skb, max_headroom);
1165                 if (!new_skb)
1166                         goto tx_err_dst_release;
1167
1168                 if (skb->sk)
1169                         skb_set_owner_w(new_skb, skb->sk);
1170                 consume_skb(skb);
1171                 skb = new_skb;
1172         }
1173
1174         if (t->parms.collect_md) {
1175                 if (t->encap.type != TUNNEL_ENCAP_NONE)
1176                         goto tx_err_dst_release;
1177         } else {
1178                 if (use_cache && ndst)
1179                         dst_cache_set_ip6(&t->dst_cache, ndst, &fl6->saddr);
1180         }
1181         skb_dst_set(skb, dst);
1182
1183         if (hop_limit == 0) {
1184                 if (skb->protocol == htons(ETH_P_IP))
1185                         hop_limit = ip_hdr(skb)->ttl;
1186                 else if (skb->protocol == htons(ETH_P_IPV6))
1187                         hop_limit = ipv6_hdr(skb)->hop_limit;
1188                 else
1189                         hop_limit = ip6_dst_hoplimit(dst);
1190         }
1191
1192         /* Calculate max headroom for all the headers and adjust
1193          * needed_headroom if necessary.
1194          */
1195         max_headroom = LL_RESERVED_SPACE(dst->dev) + sizeof(struct ipv6hdr)
1196                         + dst->header_len + t->hlen;
1197         if (max_headroom > dev->needed_headroom)
1198                 dev->needed_headroom = max_headroom;
1199
1200         err = ip6_tnl_encap(skb, t, &proto, fl6);
1201         if (err)
1202                 return err;
1203
1204         if (encap_limit >= 0) {
1205                 init_tel_txopt(&opt, encap_limit);
1206                 ipv6_push_frag_opts(skb, &opt.ops, &proto);
1207         }
1208
1209         skb_push(skb, sizeof(struct ipv6hdr));
1210         skb_reset_network_header(skb);
1211         ipv6h = ipv6_hdr(skb);
1212         ip6_flow_hdr(ipv6h, dsfield,
1213                      ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6));
1214         ipv6h->hop_limit = hop_limit;
1215         ipv6h->nexthdr = proto;
1216         ipv6h->saddr = fl6->saddr;
1217         ipv6h->daddr = fl6->daddr;
1218         ip6tunnel_xmit(NULL, skb, dev);
1219         return 0;
1220 tx_err_link_failure:
1221         stats->tx_carrier_errors++;
1222         dst_link_failure(skb);
1223 tx_err_dst_release:
1224         dst_release(dst);
1225         return err;
1226 }
1227 EXPORT_SYMBOL(ip6_tnl_xmit);
1228
1229 static inline int
1230 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1231 {
1232         struct ip6_tnl *t = netdev_priv(dev);
1233         const struct iphdr  *iph;
1234         int encap_limit = -1;
1235         struct flowi6 fl6;
1236         __u8 dsfield;
1237         __u32 mtu;
1238         u8 tproto;
1239         int err;
1240
1241         iph = ip_hdr(skb);
1242         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1243
1244         tproto = READ_ONCE(t->parms.proto);
1245         if (tproto != IPPROTO_IPIP && tproto != 0)
1246                 return -1;
1247
1248         if (t->parms.collect_md) {
1249                 struct ip_tunnel_info *tun_info;
1250                 const struct ip_tunnel_key *key;
1251
1252                 tun_info = skb_tunnel_info(skb);
1253                 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1254                              ip_tunnel_info_af(tun_info) != AF_INET6))
1255                         return -1;
1256                 key = &tun_info->key;
1257                 memset(&fl6, 0, sizeof(fl6));
1258                 fl6.flowi6_proto = IPPROTO_IPIP;
1259                 fl6.saddr = key->u.ipv6.src;
1260                 fl6.daddr = key->u.ipv6.dst;
1261                 fl6.flowlabel = key->label;
1262                 dsfield =  key->tos;
1263         } else {
1264                 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1265                         encap_limit = t->parms.encap_limit;
1266
1267                 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1268                 fl6.flowi6_proto = IPPROTO_IPIP;
1269
1270                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1271                         dsfield = ipv4_get_dsfield(iph);
1272                 else
1273                         dsfield = ip6_tclass(t->parms.flowinfo);
1274                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1275                         fl6.flowi6_mark = skb->mark;
1276                 else
1277                         fl6.flowi6_mark = t->parms.fwmark;
1278         }
1279
1280         fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1281
1282         if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1283                 return -1;
1284
1285         dsfield = INET_ECN_encapsulate(dsfield, ipv4_get_dsfield(iph));
1286
1287         skb_set_inner_ipproto(skb, IPPROTO_IPIP);
1288
1289         err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1290                            IPPROTO_IPIP);
1291         if (err != 0) {
1292                 /* XXX: send ICMP error even if DF is not set. */
1293                 if (err == -EMSGSIZE)
1294                         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
1295                                   htonl(mtu));
1296                 return -1;
1297         }
1298
1299         return 0;
1300 }
1301
1302 static inline int
1303 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1304 {
1305         struct ip6_tnl *t = netdev_priv(dev);
1306         struct ipv6hdr *ipv6h;
1307         int encap_limit = -1;
1308         __u16 offset;
1309         struct flowi6 fl6;
1310         __u8 dsfield;
1311         __u32 mtu;
1312         u8 tproto;
1313         int err;
1314
1315         ipv6h = ipv6_hdr(skb);
1316         tproto = READ_ONCE(t->parms.proto);
1317         if ((tproto != IPPROTO_IPV6 && tproto != 0) ||
1318             ip6_tnl_addr_conflict(t, ipv6h))
1319                 return -1;
1320
1321         if (t->parms.collect_md) {
1322                 struct ip_tunnel_info *tun_info;
1323                 const struct ip_tunnel_key *key;
1324
1325                 tun_info = skb_tunnel_info(skb);
1326                 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1327                              ip_tunnel_info_af(tun_info) != AF_INET6))
1328                         return -1;
1329                 key = &tun_info->key;
1330                 memset(&fl6, 0, sizeof(fl6));
1331                 fl6.flowi6_proto = IPPROTO_IPV6;
1332                 fl6.saddr = key->u.ipv6.src;
1333                 fl6.daddr = key->u.ipv6.dst;
1334                 fl6.flowlabel = key->label;
1335                 dsfield = key->tos;
1336         } else {
1337                 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
1338                 /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
1339                 ipv6h = ipv6_hdr(skb);
1340                 if (offset > 0) {
1341                         struct ipv6_tlv_tnl_enc_lim *tel;
1342
1343                         tel = (void *)&skb_network_header(skb)[offset];
1344                         if (tel->encap_limit == 0) {
1345                                 icmpv6_send(skb, ICMPV6_PARAMPROB,
1346                                             ICMPV6_HDR_FIELD, offset + 2);
1347                                 return -1;
1348                         }
1349                         encap_limit = tel->encap_limit - 1;
1350                 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
1351                         encap_limit = t->parms.encap_limit;
1352                 }
1353
1354                 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1355                 fl6.flowi6_proto = IPPROTO_IPV6;
1356
1357                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1358                         dsfield = ipv6_get_dsfield(ipv6h);
1359                 else
1360                         dsfield = ip6_tclass(t->parms.flowinfo);
1361                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1362                         fl6.flowlabel |= ip6_flowlabel(ipv6h);
1363                 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1364                         fl6.flowi6_mark = skb->mark;
1365                 else
1366                         fl6.flowi6_mark = t->parms.fwmark;
1367         }
1368
1369         fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1370
1371         if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1372                 return -1;
1373
1374         dsfield = INET_ECN_encapsulate(dsfield, ipv6_get_dsfield(ipv6h));
1375
1376         skb_set_inner_ipproto(skb, IPPROTO_IPV6);
1377
1378         err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1379                            IPPROTO_IPV6);
1380         if (err != 0) {
1381                 if (err == -EMSGSIZE)
1382                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1383                 return -1;
1384         }
1385
1386         return 0;
1387 }
1388
1389 static netdev_tx_t
1390 ip6_tnl_start_xmit(struct sk_buff *skb, struct net_device *dev)
1391 {
1392         struct ip6_tnl *t = netdev_priv(dev);
1393         struct net_device_stats *stats = &t->dev->stats;
1394         int ret;
1395
1396         if (!pskb_inet_may_pull(skb))
1397                 goto tx_err;
1398
1399         switch (skb->protocol) {
1400         case htons(ETH_P_IP):
1401                 ret = ip4ip6_tnl_xmit(skb, dev);
1402                 break;
1403         case htons(ETH_P_IPV6):
1404                 ret = ip6ip6_tnl_xmit(skb, dev);
1405                 break;
1406         default:
1407                 goto tx_err;
1408         }
1409
1410         if (ret < 0)
1411                 goto tx_err;
1412
1413         return NETDEV_TX_OK;
1414
1415 tx_err:
1416         stats->tx_errors++;
1417         stats->tx_dropped++;
1418         kfree_skb(skb);
1419         return NETDEV_TX_OK;
1420 }
1421
1422 static void ip6_tnl_link_config(struct ip6_tnl *t)
1423 {
1424         struct net_device *dev = t->dev;
1425         struct __ip6_tnl_parm *p = &t->parms;
1426         struct flowi6 *fl6 = &t->fl.u.ip6;
1427         int t_hlen;
1428
1429         memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1430         memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1431
1432         /* Set up flowi template */
1433         fl6->saddr = p->laddr;
1434         fl6->daddr = p->raddr;
1435         fl6->flowi6_oif = p->link;
1436         fl6->flowlabel = 0;
1437
1438         if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1439                 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1440         if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1441                 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1442
1443         p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1444         p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1445
1446         if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1447                 dev->flags |= IFF_POINTOPOINT;
1448         else
1449                 dev->flags &= ~IFF_POINTOPOINT;
1450
1451         t->tun_hlen = 0;
1452         t->hlen = t->encap_hlen + t->tun_hlen;
1453         t_hlen = t->hlen + sizeof(struct ipv6hdr);
1454
1455         if (p->flags & IP6_TNL_F_CAP_XMIT) {
1456                 int strict = (ipv6_addr_type(&p->raddr) &
1457                               (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1458
1459                 struct rt6_info *rt = rt6_lookup(t->net,
1460                                                  &p->raddr, &p->laddr,
1461                                                  p->link, NULL, strict);
1462
1463                 if (!rt)
1464                         return;
1465
1466                 if (rt->dst.dev) {
1467                         dev->hard_header_len = rt->dst.dev->hard_header_len +
1468                                 t_hlen;
1469
1470                         dev->mtu = rt->dst.dev->mtu - t_hlen;
1471                         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1472                                 dev->mtu -= 8;
1473
1474                         if (dev->mtu < IPV6_MIN_MTU)
1475                                 dev->mtu = IPV6_MIN_MTU;
1476                 }
1477                 ip6_rt_put(rt);
1478         }
1479 }
1480
1481 /**
1482  * ip6_tnl_change - update the tunnel parameters
1483  *   @t: tunnel to be changed
1484  *   @p: tunnel configuration parameters
1485  *
1486  * Description:
1487  *   ip6_tnl_change() updates the tunnel parameters
1488  **/
1489
1490 static int
1491 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1492 {
1493         t->parms.laddr = p->laddr;
1494         t->parms.raddr = p->raddr;
1495         t->parms.flags = p->flags;
1496         t->parms.hop_limit = p->hop_limit;
1497         t->parms.encap_limit = p->encap_limit;
1498         t->parms.flowinfo = p->flowinfo;
1499         t->parms.link = p->link;
1500         t->parms.proto = p->proto;
1501         t->parms.fwmark = p->fwmark;
1502         dst_cache_reset(&t->dst_cache);
1503         ip6_tnl_link_config(t);
1504         return 0;
1505 }
1506
1507 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1508 {
1509         struct net *net = t->net;
1510         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1511         int err;
1512
1513         ip6_tnl_unlink(ip6n, t);
1514         synchronize_net();
1515         err = ip6_tnl_change(t, p);
1516         ip6_tnl_link(ip6n, t);
1517         netdev_state_change(t->dev);
1518         return err;
1519 }
1520
1521 static int ip6_tnl0_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1522 {
1523         /* for default tnl0 device allow to change only the proto */
1524         t->parms.proto = p->proto;
1525         netdev_state_change(t->dev);
1526         return 0;
1527 }
1528
1529 static void
1530 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1531 {
1532         p->laddr = u->laddr;
1533         p->raddr = u->raddr;
1534         p->flags = u->flags;
1535         p->hop_limit = u->hop_limit;
1536         p->encap_limit = u->encap_limit;
1537         p->flowinfo = u->flowinfo;
1538         p->link = u->link;
1539         p->proto = u->proto;
1540         memcpy(p->name, u->name, sizeof(u->name));
1541 }
1542
1543 static void
1544 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1545 {
1546         u->laddr = p->laddr;
1547         u->raddr = p->raddr;
1548         u->flags = p->flags;
1549         u->hop_limit = p->hop_limit;
1550         u->encap_limit = p->encap_limit;
1551         u->flowinfo = p->flowinfo;
1552         u->link = p->link;
1553         u->proto = p->proto;
1554         memcpy(u->name, p->name, sizeof(u->name));
1555 }
1556
1557 /**
1558  * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1559  *   @dev: virtual device associated with tunnel
1560  *   @ifr: parameters passed from userspace
1561  *   @cmd: command to be performed
1562  *
1563  * Description:
1564  *   ip6_tnl_ioctl() is used for managing IPv6 tunnels
1565  *   from userspace.
1566  *
1567  *   The possible commands are the following:
1568  *     %SIOCGETTUNNEL: get tunnel parameters for device
1569  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1570  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
1571  *     %SIOCDELTUNNEL: delete tunnel
1572  *
1573  *   The fallback device "ip6tnl0", created during module
1574  *   initialization, can be used for creating other tunnel devices.
1575  *
1576  * Return:
1577  *   0 on success,
1578  *   %-EFAULT if unable to copy data to or from userspace,
1579  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
1580  *   %-EINVAL if passed tunnel parameters are invalid,
1581  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
1582  *   %-ENODEV if attempting to change or delete a nonexisting device
1583  **/
1584
1585 static int
1586 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1587 {
1588         int err = 0;
1589         struct ip6_tnl_parm p;
1590         struct __ip6_tnl_parm p1;
1591         struct ip6_tnl *t = netdev_priv(dev);
1592         struct net *net = t->net;
1593         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1594
1595         memset(&p1, 0, sizeof(p1));
1596
1597         switch (cmd) {
1598         case SIOCGETTUNNEL:
1599                 if (dev == ip6n->fb_tnl_dev) {
1600                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1601                                 err = -EFAULT;
1602                                 break;
1603                         }
1604                         ip6_tnl_parm_from_user(&p1, &p);
1605                         t = ip6_tnl_locate(net, &p1, 0);
1606                         if (IS_ERR(t))
1607                                 t = netdev_priv(dev);
1608                 } else {
1609                         memset(&p, 0, sizeof(p));
1610                 }
1611                 ip6_tnl_parm_to_user(&p, &t->parms);
1612                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) {
1613                         err = -EFAULT;
1614                 }
1615                 break;
1616         case SIOCADDTUNNEL:
1617         case SIOCCHGTUNNEL:
1618                 err = -EPERM;
1619                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1620                         break;
1621                 err = -EFAULT;
1622                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1623                         break;
1624                 err = -EINVAL;
1625                 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1626                     p.proto != 0)
1627                         break;
1628                 ip6_tnl_parm_from_user(&p1, &p);
1629                 t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1630                 if (cmd == SIOCCHGTUNNEL) {
1631                         if (!IS_ERR(t)) {
1632                                 if (t->dev != dev) {
1633                                         err = -EEXIST;
1634                                         break;
1635                                 }
1636                         } else
1637                                 t = netdev_priv(dev);
1638                         if (dev == ip6n->fb_tnl_dev)
1639                                 err = ip6_tnl0_update(t, &p1);
1640                         else
1641                                 err = ip6_tnl_update(t, &p1);
1642                 }
1643                 if (!IS_ERR(t)) {
1644                         err = 0;
1645                         ip6_tnl_parm_to_user(&p, &t->parms);
1646                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1647                                 err = -EFAULT;
1648
1649                 } else {
1650                         err = PTR_ERR(t);
1651                 }
1652                 break;
1653         case SIOCDELTUNNEL:
1654                 err = -EPERM;
1655                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1656                         break;
1657
1658                 if (dev == ip6n->fb_tnl_dev) {
1659                         err = -EFAULT;
1660                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1661                                 break;
1662                         err = -ENOENT;
1663                         ip6_tnl_parm_from_user(&p1, &p);
1664                         t = ip6_tnl_locate(net, &p1, 0);
1665                         if (IS_ERR(t))
1666                                 break;
1667                         err = -EPERM;
1668                         if (t->dev == ip6n->fb_tnl_dev)
1669                                 break;
1670                         dev = t->dev;
1671                 }
1672                 err = 0;
1673                 unregister_netdevice(dev);
1674                 break;
1675         default:
1676                 err = -EINVAL;
1677         }
1678         return err;
1679 }
1680
1681 /**
1682  * ip6_tnl_change_mtu - change mtu manually for tunnel device
1683  *   @dev: virtual device associated with tunnel
1684  *   @new_mtu: the new mtu
1685  *
1686  * Return:
1687  *   0 on success,
1688  *   %-EINVAL if mtu too small
1689  **/
1690
1691 int ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1692 {
1693         struct ip6_tnl *tnl = netdev_priv(dev);
1694
1695         if (tnl->parms.proto == IPPROTO_IPV6) {
1696                 if (new_mtu < IPV6_MIN_MTU)
1697                         return -EINVAL;
1698         } else {
1699                 if (new_mtu < ETH_MIN_MTU)
1700                         return -EINVAL;
1701         }
1702         if (tnl->parms.proto == IPPROTO_IPV6 || tnl->parms.proto == 0) {
1703                 if (new_mtu > IP6_MAX_MTU - dev->hard_header_len)
1704                         return -EINVAL;
1705         } else {
1706                 if (new_mtu > IP_MAX_MTU - dev->hard_header_len)
1707                         return -EINVAL;
1708         }
1709         dev->mtu = new_mtu;
1710         return 0;
1711 }
1712 EXPORT_SYMBOL(ip6_tnl_change_mtu);
1713
1714 int ip6_tnl_get_iflink(const struct net_device *dev)
1715 {
1716         struct ip6_tnl *t = netdev_priv(dev);
1717
1718         return t->parms.link;
1719 }
1720 EXPORT_SYMBOL(ip6_tnl_get_iflink);
1721
1722 int ip6_tnl_encap_add_ops(const struct ip6_tnl_encap_ops *ops,
1723                           unsigned int num)
1724 {
1725         if (num >= MAX_IPTUN_ENCAP_OPS)
1726                 return -ERANGE;
1727
1728         return !cmpxchg((const struct ip6_tnl_encap_ops **)
1729                         &ip6tun_encaps[num],
1730                         NULL, ops) ? 0 : -1;
1731 }
1732 EXPORT_SYMBOL(ip6_tnl_encap_add_ops);
1733
1734 int ip6_tnl_encap_del_ops(const struct ip6_tnl_encap_ops *ops,
1735                           unsigned int num)
1736 {
1737         int ret;
1738
1739         if (num >= MAX_IPTUN_ENCAP_OPS)
1740                 return -ERANGE;
1741
1742         ret = (cmpxchg((const struct ip6_tnl_encap_ops **)
1743                        &ip6tun_encaps[num],
1744                        ops, NULL) == ops) ? 0 : -1;
1745
1746         synchronize_net();
1747
1748         return ret;
1749 }
1750 EXPORT_SYMBOL(ip6_tnl_encap_del_ops);
1751
1752 int ip6_tnl_encap_setup(struct ip6_tnl *t,
1753                         struct ip_tunnel_encap *ipencap)
1754 {
1755         int hlen;
1756
1757         memset(&t->encap, 0, sizeof(t->encap));
1758
1759         hlen = ip6_encap_hlen(ipencap);
1760         if (hlen < 0)
1761                 return hlen;
1762
1763         t->encap.type = ipencap->type;
1764         t->encap.sport = ipencap->sport;
1765         t->encap.dport = ipencap->dport;
1766         t->encap.flags = ipencap->flags;
1767
1768         t->encap_hlen = hlen;
1769         t->hlen = t->encap_hlen + t->tun_hlen;
1770
1771         return 0;
1772 }
1773 EXPORT_SYMBOL_GPL(ip6_tnl_encap_setup);
1774
1775 static const struct net_device_ops ip6_tnl_netdev_ops = {
1776         .ndo_init       = ip6_tnl_dev_init,
1777         .ndo_uninit     = ip6_tnl_dev_uninit,
1778         .ndo_start_xmit = ip6_tnl_start_xmit,
1779         .ndo_do_ioctl   = ip6_tnl_ioctl,
1780         .ndo_change_mtu = ip6_tnl_change_mtu,
1781         .ndo_get_stats  = ip6_get_stats,
1782         .ndo_get_iflink = ip6_tnl_get_iflink,
1783 };
1784
1785 #define IPXIPX_FEATURES (NETIF_F_SG |           \
1786                          NETIF_F_FRAGLIST |     \
1787                          NETIF_F_HIGHDMA |      \
1788                          NETIF_F_GSO_SOFTWARE | \
1789                          NETIF_F_HW_CSUM)
1790
1791 /**
1792  * ip6_tnl_dev_setup - setup virtual tunnel device
1793  *   @dev: virtual device associated with tunnel
1794  *
1795  * Description:
1796  *   Initialize function pointers and device parameters
1797  **/
1798
1799 static void ip6_tnl_dev_setup(struct net_device *dev)
1800 {
1801         dev->netdev_ops = &ip6_tnl_netdev_ops;
1802         dev->needs_free_netdev = true;
1803         dev->priv_destructor = ip6_dev_free;
1804
1805         dev->type = ARPHRD_TUNNEL6;
1806         dev->flags |= IFF_NOARP;
1807         dev->addr_len = sizeof(struct in6_addr);
1808         dev->features |= NETIF_F_LLTX;
1809         netif_keep_dst(dev);
1810
1811         dev->features           |= IPXIPX_FEATURES;
1812         dev->hw_features        |= IPXIPX_FEATURES;
1813
1814         /* This perm addr will be used as interface identifier by IPv6 */
1815         dev->addr_assign_type = NET_ADDR_RANDOM;
1816         eth_random_addr(dev->perm_addr);
1817 }
1818
1819
1820 /**
1821  * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1822  *   @dev: virtual device associated with tunnel
1823  **/
1824
1825 static inline int
1826 ip6_tnl_dev_init_gen(struct net_device *dev)
1827 {
1828         struct ip6_tnl *t = netdev_priv(dev);
1829         int ret;
1830         int t_hlen;
1831
1832         t->dev = dev;
1833         t->net = dev_net(dev);
1834         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1835         if (!dev->tstats)
1836                 return -ENOMEM;
1837
1838         ret = dst_cache_init(&t->dst_cache, GFP_KERNEL);
1839         if (ret)
1840                 goto free_stats;
1841
1842         ret = gro_cells_init(&t->gro_cells, dev);
1843         if (ret)
1844                 goto destroy_dst;
1845
1846         t->tun_hlen = 0;
1847         t->hlen = t->encap_hlen + t->tun_hlen;
1848         t_hlen = t->hlen + sizeof(struct ipv6hdr);
1849
1850         dev->type = ARPHRD_TUNNEL6;
1851         dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1852         dev->mtu = ETH_DATA_LEN - t_hlen;
1853         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1854                 dev->mtu -= 8;
1855         dev->min_mtu = ETH_MIN_MTU;
1856         dev->max_mtu = IP6_MAX_MTU - dev->hard_header_len;
1857
1858         return 0;
1859
1860 destroy_dst:
1861         dst_cache_destroy(&t->dst_cache);
1862 free_stats:
1863         free_percpu(dev->tstats);
1864         dev->tstats = NULL;
1865
1866         return ret;
1867 }
1868
1869 /**
1870  * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1871  *   @dev: virtual device associated with tunnel
1872  **/
1873
1874 static int ip6_tnl_dev_init(struct net_device *dev)
1875 {
1876         struct ip6_tnl *t = netdev_priv(dev);
1877         int err = ip6_tnl_dev_init_gen(dev);
1878
1879         if (err)
1880                 return err;
1881         ip6_tnl_link_config(t);
1882         if (t->parms.collect_md) {
1883                 dev->features |= NETIF_F_NETNS_LOCAL;
1884                 netif_keep_dst(dev);
1885         }
1886         return 0;
1887 }
1888
1889 /**
1890  * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1891  *   @dev: fallback device
1892  *
1893  * Return: 0
1894  **/
1895
1896 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1897 {
1898         struct ip6_tnl *t = netdev_priv(dev);
1899         struct net *net = dev_net(dev);
1900         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1901
1902         t->parms.proto = IPPROTO_IPV6;
1903         dev_hold(dev);
1904
1905         rcu_assign_pointer(ip6n->tnls_wc[0], t);
1906         return 0;
1907 }
1908
1909 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[],
1910                             struct netlink_ext_ack *extack)
1911 {
1912         u8 proto;
1913
1914         if (!data || !data[IFLA_IPTUN_PROTO])
1915                 return 0;
1916
1917         proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1918         if (proto != IPPROTO_IPV6 &&
1919             proto != IPPROTO_IPIP &&
1920             proto != 0)
1921                 return -EINVAL;
1922
1923         return 0;
1924 }
1925
1926 static void ip6_tnl_netlink_parms(struct nlattr *data[],
1927                                   struct __ip6_tnl_parm *parms)
1928 {
1929         memset(parms, 0, sizeof(*parms));
1930
1931         if (!data)
1932                 return;
1933
1934         if (data[IFLA_IPTUN_LINK])
1935                 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1936
1937         if (data[IFLA_IPTUN_LOCAL])
1938                 parms->laddr = nla_get_in6_addr(data[IFLA_IPTUN_LOCAL]);
1939
1940         if (data[IFLA_IPTUN_REMOTE])
1941                 parms->raddr = nla_get_in6_addr(data[IFLA_IPTUN_REMOTE]);
1942
1943         if (data[IFLA_IPTUN_TTL])
1944                 parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
1945
1946         if (data[IFLA_IPTUN_ENCAP_LIMIT])
1947                 parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
1948
1949         if (data[IFLA_IPTUN_FLOWINFO])
1950                 parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
1951
1952         if (data[IFLA_IPTUN_FLAGS])
1953                 parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
1954
1955         if (data[IFLA_IPTUN_PROTO])
1956                 parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1957
1958         if (data[IFLA_IPTUN_COLLECT_METADATA])
1959                 parms->collect_md = true;
1960
1961         if (data[IFLA_IPTUN_FWMARK])
1962                 parms->fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
1963 }
1964
1965 static bool ip6_tnl_netlink_encap_parms(struct nlattr *data[],
1966                                         struct ip_tunnel_encap *ipencap)
1967 {
1968         bool ret = false;
1969
1970         memset(ipencap, 0, sizeof(*ipencap));
1971
1972         if (!data)
1973                 return ret;
1974
1975         if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1976                 ret = true;
1977                 ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1978         }
1979
1980         if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1981                 ret = true;
1982                 ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1983         }
1984
1985         if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1986                 ret = true;
1987                 ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1988         }
1989
1990         if (data[IFLA_IPTUN_ENCAP_DPORT]) {
1991                 ret = true;
1992                 ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
1993         }
1994
1995         return ret;
1996 }
1997
1998 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev,
1999                            struct nlattr *tb[], struct nlattr *data[],
2000                            struct netlink_ext_ack *extack)
2001 {
2002         struct net *net = dev_net(dev);
2003         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2004         struct ip_tunnel_encap ipencap;
2005         struct ip6_tnl *nt, *t;
2006         int err;
2007
2008         nt = netdev_priv(dev);
2009
2010         if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
2011                 err = ip6_tnl_encap_setup(nt, &ipencap);
2012                 if (err < 0)
2013                         return err;
2014         }
2015
2016         ip6_tnl_netlink_parms(data, &nt->parms);
2017
2018         if (nt->parms.collect_md) {
2019                 if (rtnl_dereference(ip6n->collect_md_tun))
2020                         return -EEXIST;
2021         } else {
2022                 t = ip6_tnl_locate(net, &nt->parms, 0);
2023                 if (!IS_ERR(t))
2024                         return -EEXIST;
2025         }
2026
2027         err = ip6_tnl_create2(dev);
2028         if (!err && tb[IFLA_MTU])
2029                 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
2030
2031         return err;
2032 }
2033
2034 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
2035                               struct nlattr *data[],
2036                               struct netlink_ext_ack *extack)
2037 {
2038         struct ip6_tnl *t = netdev_priv(dev);
2039         struct __ip6_tnl_parm p;
2040         struct net *net = t->net;
2041         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2042         struct ip_tunnel_encap ipencap;
2043
2044         if (dev == ip6n->fb_tnl_dev)
2045                 return -EINVAL;
2046
2047         if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
2048                 int err = ip6_tnl_encap_setup(t, &ipencap);
2049
2050                 if (err < 0)
2051                         return err;
2052         }
2053         ip6_tnl_netlink_parms(data, &p);
2054         if (p.collect_md)
2055                 return -EINVAL;
2056
2057         t = ip6_tnl_locate(net, &p, 0);
2058         if (!IS_ERR(t)) {
2059                 if (t->dev != dev)
2060                         return -EEXIST;
2061         } else
2062                 t = netdev_priv(dev);
2063
2064         return ip6_tnl_update(t, &p);
2065 }
2066
2067 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head)
2068 {
2069         struct net *net = dev_net(dev);
2070         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2071
2072         if (dev != ip6n->fb_tnl_dev)
2073                 unregister_netdevice_queue(dev, head);
2074 }
2075
2076 static size_t ip6_tnl_get_size(const struct net_device *dev)
2077 {
2078         return
2079                 /* IFLA_IPTUN_LINK */
2080                 nla_total_size(4) +
2081                 /* IFLA_IPTUN_LOCAL */
2082                 nla_total_size(sizeof(struct in6_addr)) +
2083                 /* IFLA_IPTUN_REMOTE */
2084                 nla_total_size(sizeof(struct in6_addr)) +
2085                 /* IFLA_IPTUN_TTL */
2086                 nla_total_size(1) +
2087                 /* IFLA_IPTUN_ENCAP_LIMIT */
2088                 nla_total_size(1) +
2089                 /* IFLA_IPTUN_FLOWINFO */
2090                 nla_total_size(4) +
2091                 /* IFLA_IPTUN_FLAGS */
2092                 nla_total_size(4) +
2093                 /* IFLA_IPTUN_PROTO */
2094                 nla_total_size(1) +
2095                 /* IFLA_IPTUN_ENCAP_TYPE */
2096                 nla_total_size(2) +
2097                 /* IFLA_IPTUN_ENCAP_FLAGS */
2098                 nla_total_size(2) +
2099                 /* IFLA_IPTUN_ENCAP_SPORT */
2100                 nla_total_size(2) +
2101                 /* IFLA_IPTUN_ENCAP_DPORT */
2102                 nla_total_size(2) +
2103                 /* IFLA_IPTUN_COLLECT_METADATA */
2104                 nla_total_size(0) +
2105                 /* IFLA_IPTUN_FWMARK */
2106                 nla_total_size(4) +
2107                 0;
2108 }
2109
2110 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
2111 {
2112         struct ip6_tnl *tunnel = netdev_priv(dev);
2113         struct __ip6_tnl_parm *parm = &tunnel->parms;
2114
2115         if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
2116             nla_put_in6_addr(skb, IFLA_IPTUN_LOCAL, &parm->laddr) ||
2117             nla_put_in6_addr(skb, IFLA_IPTUN_REMOTE, &parm->raddr) ||
2118             nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
2119             nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
2120             nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
2121             nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
2122             nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto) ||
2123             nla_put_u32(skb, IFLA_IPTUN_FWMARK, parm->fwmark))
2124                 goto nla_put_failure;
2125
2126         if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE, tunnel->encap.type) ||
2127             nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT, tunnel->encap.sport) ||
2128             nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT, tunnel->encap.dport) ||
2129             nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS, tunnel->encap.flags))
2130                 goto nla_put_failure;
2131
2132         if (parm->collect_md)
2133                 if (nla_put_flag(skb, IFLA_IPTUN_COLLECT_METADATA))
2134                         goto nla_put_failure;
2135
2136         return 0;
2137
2138 nla_put_failure:
2139         return -EMSGSIZE;
2140 }
2141
2142 struct net *ip6_tnl_get_link_net(const struct net_device *dev)
2143 {
2144         struct ip6_tnl *tunnel = netdev_priv(dev);
2145
2146         return tunnel->net;
2147 }
2148 EXPORT_SYMBOL(ip6_tnl_get_link_net);
2149
2150 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
2151         [IFLA_IPTUN_LINK]               = { .type = NLA_U32 },
2152         [IFLA_IPTUN_LOCAL]              = { .len = sizeof(struct in6_addr) },
2153         [IFLA_IPTUN_REMOTE]             = { .len = sizeof(struct in6_addr) },
2154         [IFLA_IPTUN_TTL]                = { .type = NLA_U8 },
2155         [IFLA_IPTUN_ENCAP_LIMIT]        = { .type = NLA_U8 },
2156         [IFLA_IPTUN_FLOWINFO]           = { .type = NLA_U32 },
2157         [IFLA_IPTUN_FLAGS]              = { .type = NLA_U32 },
2158         [IFLA_IPTUN_PROTO]              = { .type = NLA_U8 },
2159         [IFLA_IPTUN_ENCAP_TYPE]         = { .type = NLA_U16 },
2160         [IFLA_IPTUN_ENCAP_FLAGS]        = { .type = NLA_U16 },
2161         [IFLA_IPTUN_ENCAP_SPORT]        = { .type = NLA_U16 },
2162         [IFLA_IPTUN_ENCAP_DPORT]        = { .type = NLA_U16 },
2163         [IFLA_IPTUN_COLLECT_METADATA]   = { .type = NLA_FLAG },
2164         [IFLA_IPTUN_FWMARK]             = { .type = NLA_U32 },
2165 };
2166
2167 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
2168         .kind           = "ip6tnl",
2169         .maxtype        = IFLA_IPTUN_MAX,
2170         .policy         = ip6_tnl_policy,
2171         .priv_size      = sizeof(struct ip6_tnl),
2172         .setup          = ip6_tnl_dev_setup,
2173         .validate       = ip6_tnl_validate,
2174         .newlink        = ip6_tnl_newlink,
2175         .changelink     = ip6_tnl_changelink,
2176         .dellink        = ip6_tnl_dellink,
2177         .get_size       = ip6_tnl_get_size,
2178         .fill_info      = ip6_tnl_fill_info,
2179         .get_link_net   = ip6_tnl_get_link_net,
2180 };
2181
2182 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
2183         .handler        = ip4ip6_rcv,
2184         .err_handler    = ip4ip6_err,
2185         .priority       =       1,
2186 };
2187
2188 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
2189         .handler        = ip6ip6_rcv,
2190         .err_handler    = ip6ip6_err,
2191         .priority       =       1,
2192 };
2193
2194 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net, struct list_head *list)
2195 {
2196         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2197         struct net_device *dev, *aux;
2198         int h;
2199         struct ip6_tnl *t;
2200
2201         for_each_netdev_safe(net, dev, aux)
2202                 if (dev->rtnl_link_ops == &ip6_link_ops)
2203                         unregister_netdevice_queue(dev, list);
2204
2205         for (h = 0; h < IP6_TUNNEL_HASH_SIZE; h++) {
2206                 t = rtnl_dereference(ip6n->tnls_r_l[h]);
2207                 while (t) {
2208                         /* If dev is in the same netns, it has already
2209                          * been added to the list by the previous loop.
2210                          */
2211                         if (!net_eq(dev_net(t->dev), net))
2212                                 unregister_netdevice_queue(t->dev, list);
2213                         t = rtnl_dereference(t->next);
2214                 }
2215         }
2216 }
2217
2218 static int __net_init ip6_tnl_init_net(struct net *net)
2219 {
2220         struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2221         struct ip6_tnl *t = NULL;
2222         int err;
2223
2224         ip6n->tnls[0] = ip6n->tnls_wc;
2225         ip6n->tnls[1] = ip6n->tnls_r_l;
2226
2227         if (!net_has_fallback_tunnels(net))
2228                 return 0;
2229         err = -ENOMEM;
2230         ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
2231                                         NET_NAME_UNKNOWN, ip6_tnl_dev_setup);
2232
2233         if (!ip6n->fb_tnl_dev)
2234                 goto err_alloc_dev;
2235         dev_net_set(ip6n->fb_tnl_dev, net);
2236         ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops;
2237         /* FB netdevice is special: we have one, and only one per netns.
2238          * Allowing to move it to another netns is clearly unsafe.
2239          */
2240         ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL;
2241
2242         err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
2243         if (err < 0)
2244                 goto err_register;
2245
2246         err = register_netdev(ip6n->fb_tnl_dev);
2247         if (err < 0)
2248                 goto err_register;
2249
2250         t = netdev_priv(ip6n->fb_tnl_dev);
2251
2252         strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
2253         return 0;
2254
2255 err_register:
2256         free_netdev(ip6n->fb_tnl_dev);
2257 err_alloc_dev:
2258         return err;
2259 }
2260
2261 static void __net_exit ip6_tnl_exit_batch_net(struct list_head *net_list)
2262 {
2263         struct net *net;
2264         LIST_HEAD(list);
2265
2266         rtnl_lock();
2267         list_for_each_entry(net, net_list, exit_list)
2268                 ip6_tnl_destroy_tunnels(net, &list);
2269         unregister_netdevice_many(&list);
2270         rtnl_unlock();
2271 }
2272
2273 static struct pernet_operations ip6_tnl_net_ops = {
2274         .init = ip6_tnl_init_net,
2275         .exit_batch = ip6_tnl_exit_batch_net,
2276         .id   = &ip6_tnl_net_id,
2277         .size = sizeof(struct ip6_tnl_net),
2278 };
2279
2280 /**
2281  * ip6_tunnel_init - register protocol and reserve needed resources
2282  *
2283  * Return: 0 on success
2284  **/
2285
2286 static int __init ip6_tunnel_init(void)
2287 {
2288         int  err;
2289
2290         if (!ipv6_mod_enabled())
2291                 return -EOPNOTSUPP;
2292
2293         err = register_pernet_device(&ip6_tnl_net_ops);
2294         if (err < 0)
2295                 goto out_pernet;
2296
2297         err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
2298         if (err < 0) {
2299                 pr_err("%s: can't register ip4ip6\n", __func__);
2300                 goto out_ip4ip6;
2301         }
2302
2303         err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
2304         if (err < 0) {
2305                 pr_err("%s: can't register ip6ip6\n", __func__);
2306                 goto out_ip6ip6;
2307         }
2308         err = rtnl_link_register(&ip6_link_ops);
2309         if (err < 0)
2310                 goto rtnl_link_failed;
2311
2312         return 0;
2313
2314 rtnl_link_failed:
2315         xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
2316 out_ip6ip6:
2317         xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
2318 out_ip4ip6:
2319         unregister_pernet_device(&ip6_tnl_net_ops);
2320 out_pernet:
2321         return err;
2322 }
2323
2324 /**
2325  * ip6_tunnel_cleanup - free resources and unregister protocol
2326  **/
2327
2328 static void __exit ip6_tunnel_cleanup(void)
2329 {
2330         rtnl_link_unregister(&ip6_link_ops);
2331         if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
2332                 pr_info("%s: can't deregister ip4ip6\n", __func__);
2333
2334         if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
2335                 pr_info("%s: can't deregister ip6ip6\n", __func__);
2336
2337         unregister_pernet_device(&ip6_tnl_net_ops);
2338 }
2339
2340 module_init(ip6_tunnel_init);
2341 module_exit(ip6_tunnel_cleanup);